Scholarlink

Due to the fact that liver cancer is diagnosed in an undesirable late stage, its early detection has been a main problem associated with using medical images for discerning between different stages of cancer. This study is aimed at solving this life-saving problem by transferring the methods of deep learning to assist in accurate and early diagnosis of liver cancer. We assess three existing deep learning models. Inception V3, VGG19 and VGG16 by applying 2350 images regarding different levels of liver cancer. These models are then trained and tested to detect whether an image is cancerous or benign based on their classification_image. Showing the best accuracy of 96.56%, precession and recall rate Inception V3 is the most reliable model for detecting liver cancer in this study. The VGG19 is closely followed by the 93.40% accuracy attained on the same dataset with VGG16, despite a simpler architectural approach. These findings demonstrate the increased accuracy in diagnostics that use Inception V3 due to its lack of default bias. These models have the potential to greatly improve early detection and treatment planning of liver cancer, resulting in better patient outcomes as well as promising further exploration into new medical imaging technologies.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Experimental Evaluation of Deep Learning Models for Liver Lesion Analysis Across Imaging Modalities

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Jagendra Singh,
  • Rajnish Kumar Chaturvedi,
  • Senthilkumar Jagatheesan,
  • Rajashree Chakraborty,
  • Jagadish S. Jakati,
  • Wilson Hrangkhawl,
  • Ishaan Singh

摘要

Due to the fact that liver cancer is diagnosed in an undesirable late stage, its early detection has been a main problem associated with using medical images for discerning between different stages of cancer. This study is aimed at solving this life-saving problem by transferring the methods of deep learning to assist in accurate and early diagnosis of liver cancer. We assess three existing deep learning models. Inception V3, VGG19 and VGG16 by applying 2350 images regarding different levels of liver cancer. These models are then trained and tested to detect whether an image is cancerous or benign based on their classification_image. Showing the best accuracy of 96.56%, precession and recall rate Inception V3 is the most reliable model for detecting liver cancer in this study. The VGG19 is closely followed by the 93.40% accuracy attained on the same dataset with VGG16, despite a simpler architectural approach. These findings demonstrate the increased accuracy in diagnostics that use Inception V3 due to its lack of default bias. These models have the potential to greatly improve early detection and treatment planning of liver cancer, resulting in better patient outcomes as well as promising further exploration into new medical imaging technologies.